Simultaneous Segmentation of
Multiple Objects among Multiple Images

Wen-Sheng Chu†, Chia-Ping Chen‡¶, Chu-Song Chen†¶

† Research Center for Information Technology Innovation, Academia Sinica, Taipei 115, Taiwan
‡ Institute of Information Science, Academia Sinica, Taipei 115, Taiwan
¶ Dept. of CSIE, National Taiwan University, Taipei 106, Taiwan

    Release Notes

  • [May 5, 2012] Executable online.
  • [Nov 12, 2010] Paper, presentation and supplementary online.
  • [Sep 24, 2010] First release of dataset and groundrturh.


  • We introduce a new cosegmentation approach, MOMI-cosegmentation, to segment multiple objects that repeatedly appear among multiple images. The proposed approach tackles a more general problem than conventional cosegmentation methods in several aspects:
    1. We require no prior knowledge about about what/how many the common objects are, nor how many times each object appears in an image.
    2. We accept a small image set consisting of more than a pair of images.
    3. Our method can process full resolution images.

  • The key idea of MOMI-cosegmentation is to incorporate a common pattern discovery algorithm with the proposed Gibbs energy model in a Markov random field framework. Our approach builds upon an observation that the detected common patterns provide useful information for estimating foreground statistics, while background statistics can be estimated from the remaining pixels. The initialization and segmentation processes of MOMI-cosegmentation are completely automatic, while the segmentation errors can be substantially reduced at the same time. Experimental results demonstrate the effectiveness of the proposed approach over state-of-the-art cosegmentation method.


  • Executable & example: [ download (zip 2.1 MB) ] [ readme (pdf 740 KB) ]
  • Dataset and groundtruth: [ download (zip 11.3 MB) ]
  • Our results: [ download (zip 7.9 MB) ]


  1. MOMI-Cosegmentation: Simultaneous Segmentation of Multiple Objects among Multiple Images
    Wen-Sheng Chu, Chia-Ping Chen and Chu-Song Chen
    Asian Conference on Computer Vision (ACCV) 2010.
    [ pdf ] [ presentation ] [ supplementary (15.3 MB) ] [ bibtex ]

  2. An Efficient Algorithm for Co-segmentation
    Dorit S. Hochbaum and Vikas Singh
    ICCV 2010.


  • These datasets are collected from Flickr and Google image search without permission from the original copyright holders. By downloading these files, you agree not to hold the authors or Academia Sinica liable for any damage, lawsuits, or other loss resulting from the possession or use of files. If you are the copyright owner of one of these images and would like it removed from the dataset, please contact Wen-Sheng Chu.


  • Wen-Sheng Chu (wensheng.chu[at]
  • Please also visit my webpage for a newer update. :)